Powder Flow Down A Vertical Pipe: The Effect of Air Flow

c 2002 Cambridge University Press
J. Fluid Mech. (2002), vol. 459, pp. 317–345. 317

DOI: 10.1017/S0022112002008042 Printed in the United Kingdom
Powder flow down a vertical pipe:

the effect of air flow
By Y. B E R T H O1 , F. G I O R G I U T T I-D A U P H I NÉ1 ,
T. R A A F A T1,2 , E. J. H I N C H3 , H. J. H E R R M A N N2
A N D J. P. H U L I N1
1
Laboratoire FAST, UMR 7608, Bat. 502, Campus Universitaire, 91405 Orsay Cedex, France
2
Laboratoire PMMH, ESPCI, UMR 7636, 10 rue Vauquelin, 75231 Paris, Cedex 05, France
3
Department of Applied Mathematics and Theoretical Physics, University of Cambridge,
Silver Street, Cambridge CB3 9EW, UK
(Received 2 March 2001 and in revised form 16 November 2001)
The dynamics of dry granular flows down a vertical glass pipe of small diameter have
been studied experimentally. Simultaneous measurements of pressure profiles, air and
grain flow rates and volume fractions of particles have been realized together with
spatio-temporal diagrams of the grain distribution down the tube. At large grain flow
rates, one observes a stationary flow characterized by high particle velocities, low
particle fractions and a downflow of air resulting in an underpressure in the upper
part of the pipe. A simple model assuming a free fall of the particles slowed down by
air friction and taking into account finite particle fraction effects through Richardson–
Zaki’s law has been developed: it reproduces pressure and particle fraction variations
with distance and estimates friction forces with the wall. At lower flow rates, sequences
of high-density plugs separated by low-density bubbles moving down at a constant
velocity are observed. The pressure is larger than outside the tube and its gradient
reflects closely the weight of the grains. Writing mass and momentum conservation
equations for the air and for the grains allows one to estimate the wall friction, which
is less than 10% of the weight for grains with a clean smooth surface but up to
30% for grains with a rougher surface. At lower flow rates, oscillating-wave regimes
resulting in large pressure fluctuations are observed and their frequency is predicted.
1. Introduction
Dry granular media are at present a very active domain of research both because
of their fundamental interest in relation to the modelling of disordered and dispersed
media and because of their many areas of practical application. Granular media
are encountered in many areas such as the food industry, and civil and chemical
engineering (Jaeger & Nagel 1992; Duran 1997; Behringer & Jenkins 1997). A
particularly important problem is the flow of these materials in pipes and channels
(Jackson 2000).
A first type of study deals with systems in which the interaction between the flowing
grains and air is negligible (Savage 1979). These flows frequently undergo instabilities
(Wang, Jackson & Sundaresan 1997) and density waves may then appear. Such waves
were for instance observed experimentally between parallel walls by Reydellet, Rioual
& Clement (2000) and can be reproduced by numerical simulations using molecular
dynamics (Pöschel 1994; Lee 1994) or lattice-gas methods (Peng & Herrmann 1994).
318 Y. Bertho and others
Analysis in terms of kinetic wave theories can be applied to these phenomena (Leibig
1994; Lee & Leibig 1994) which bear strong similarities with density waves in traffic
flows (Lighthill & Whitham 1955; Musha & Higuichi 1976).
In most cases of interest, however, the interaction of the air with the flowing grains
plays a key part. This is particularly the case in practical applications to pneumatic
transport, catalytic cracking, fluid beds and emptying storage silos. Very diverse flow
regimes such as moving packed particle beds, fluidized suspensions and combinations
of these regimes are observed in these systems (Leung & Jones 1978). These regimes
may coexist in a flow system, particularly when gas is injected locally into the flow
pipe (Mountziaris & Jackson 1990). Furthermore, these flows may display unwanted
intermittency, oscillations and/or blockages: a periodic intermittent flow of this type
is observed in ‘ticking hourglass’ experiments as grains flow through a constriction
connecting two glass reservoirs (Le Pennec et al. 1995, 1996, 1998). This effect again
results from interactions between the interstitial air and the grains.
Very detailed and elaborate theoretical models have been developed to analyse
these flows and take into account the interaction between air and the grains (Leung
& Jones 1978; Ginestra, Rangachari & Jackson 1980; Chen, Rangachari & Jackson
1984; Sinclair & Jackson 1989; Nott & Jackson 1992; Jackson 2000). These models
predict flow configurations and characteristic parameters in the different parts of the
flow from a small number of input variables (the grain flow rate for instance). When
experiments are available, the predictions of these theoretical models are generally
checked against pressure profile measurements Chen et al. (1984); other parameters are
estimated theoretically. More recent experimental work has included a larger number
of measurements, allowing one in particular to analyse the particle distribution in
the flow channels (Srivastava et al. 1998). Although the above approach is very well
suited to modelling complete large systems, it requires the introduction of a large
number of relations between the various variables, based on published literature and
on experimental or empirical correlations. This may result in significant uncertainties
in the values of key variables.
In the present work we used a different approach and performed a set of experiments
on a very simple flow system (a plain vertical glass tube with a feeding hopper at
the top and a variable constriction at the bottom) equipped with a large number of
sensors. They allowed us to determine simultaneously the grain and air flow rates, the
local solid volume fraction at two positions along the tube and the pressure profile.
In addition, a linear camera gives spatio-temporal diagrams of the grain distribution
along the tube allowing a detailed analysis of the dynamical properties of the flow.
In this way, many variables are directly measured instead of needing to be estimated
through theoretical or empirical relations. Relatively simple theoretical developments
are then needed since most other parameters of interest can generally be determined
reliably through simple conservation equations: friction forces on the pipe wall can
in particular be estimated simply.
Using a simple set-up also allows us to explore systematically a broad range of
control parameter values: as shown in previous papers (Raafat, Hulin & Herrmann
1996; Aider et al. 1999), it is possible to observe a large variety of flow regimes
by varying the grain flow rate. We observe in particular, at intermediate grain flow
rates, density waves in which compact plugs are separated by bubbles with a low
volume fraction of particles (the volume fraction of particles will in the following
be called the ‘particle fraction’). The plugs move generally at constant velocity but
may display oscillations in a few cases. At higher flow rates, a low-particle-fraction
high-velocity free-fall regime is observed. At lower flow rates, the grain flow may be
Powder flow down a vertical pipe 319
stationary or intermittent and is characterized by a low velocity and a high particle
fraction (≈ 60%): this latter regime is similar to the flows of particle beds discussed
by Ginestra et al. (1980). Its properties, and particularly the intermittency associated
with it, will be reported in a subsequent work and the present paper deals with the
two other regimes in the case of vertical tubes. As stated in the introduction, other
flow regimes can appear in such systems such as the coexistence of a free fall and
a compact regime (Mountziaris & Jackson 1990) but they were not observed in the
present work.
After describing the experimental procedure, the high-velocity free-fall regime will
be first studied experimentally and analysed theoretically: it will be shown that both
the pressure and the particle fraction profiles can be reproduced quantitatively by a
simple set of equations and that the friction force on the walls can be determined
reliably. A similar approach will be used to study the pressure distribution and the
wall friction in the constant-velocity wave regime, with particular attention given to
the dependence of wall friction on the roughness of the grain surface. The same
approach is, finally, extended to an oscillating-wave regime in which analogies with
sound propagation in two-phase flows appear.
2. Experimental procedure
2.1. Experimental set-up
In the present work, we study the flow of glass beads of diameter 2a = 175 ± 25 µm
and of density ρg = (2.5±0.02)103 kg m−3 (the glass density is measured by comparing
the weight of a picnometer of precisely known volume filled first with water and then
with a known mass of beads saturated with water). The experimental flow channel
(figure 1a) is a vertical glass pipe of length L = 1.25 m and of internal diameter
2R = 3 ± 0.05 mm welded to the bottom of a spherical hopper. The bottom end
of the pipe is fitted with a variable constriction allowing one to adjust the outflow.
Computer-controlled scales placed below the tube monitor the mass of grains flowing
out at the bottom end. The mass flow rate Qm is then determined by differentiation
with respect to time. In the following the grain flow rate is characterized by the
superficial velocity q (i.e. the volume flow rate of the grains per unit area) with
Qm
q= (1)
πR 2 ρg
The hopper is a container with one opening at its bottom serving as a connection
to the experimental tube and a side inlet allowing an inflow of air monitored by
a Honeywell gas flow rate sensor. The measurement ranges are 1 l mn−1 or 4 l mn−1
and the sensors are calibrated by means of a variable-flow-rate displacement syringe
pump; the volume flow rate Qmeas of inflowing air is measured in this way with a
precision of ± 1%. As for grains, the flow of air through the tube is characterized by
a superficial velocity qa (i.e. the volume flow rate per unit area in the experimental
tube section). Note that qa is not equal to Qmeas /πR 2 but that

1 Qm Qmeas
qa = Qmeas − = − q. (2)
πR 2 ρg πR 2
The inflow of air must compensate both the volume flow rate Qm /ρg of the grains
which flow out of the hopper and that of the air which flows out of the hopper into
the tube.
(a) (b) (c)
Pressure sensor
Air inlet Filling
and flow inlet
rate sensor
Hopper
Linear lp
Capacitance camera Plug
particle
fraction lb Bubble
sensors
lp Plug
lpb
lb Bubble
Experimental
tube Pressure
sensors
Long
lbp
Computer- bottom
controlled Adjustable z plug
scales constriction
Figure 1. (a) Experimental set-up for vertical granular flow studies. (b) Schematic view of the
free-fall regime. (c) Schematic view of the density-wave regime.
Pressure variations along the pipe are monitored by four sensors located along
the tube at vertical distances of 200, 450, 700 and 950 mm below the outlet of the
hopper. We use solid-state strain-gauge sensors attached directly to the tube wall
with a measurement range of 3 × 104 Pa, a resolution of the order of 100 Pa and a
response time of a few 10−2 s: the transducers have a low internal volume of a few
mm3 and are connected by short (5 mm long), 1 mm ID tubes to 0.5 mm diameter
holes in the experimental pipe. A fine-meshed grid is stretched across them in order
to avoid grains penetrating the sensors. Another transducer measures pressure in the
hopper.
Variations of the mean particle fraction in a tube section are estimated from the
measurement of the electrical capacitance between two 3 mm diameter cylindrical
electrodes with their ends against the outside tube wall and facing each other. Careful
shielding reduces the global capacitance of the probe to a few 10−2 pF and the
electrode assembly can be moved along the tube. Using a General Radio 1615A
capacitance bridge and a lock-in detector with a 3 ms time constant and a 10 kHz
measurement frequency gives a very good measurement stability and a noise level
equivalent to a few 10−5 pF. The DC output is assumed to vary roughly linearly
with the particle fraction since the corresponding capacitance variations are only a
minor fraction of the global value; the measurement is then calibrated by comparing
readings obtained with an empty tube and with the same tube containing a bead
packing of particle fraction 63% (estimated by weighing the grains and independently
measuring their density). The noise level corresponds to particle fraction variations
of order 0.5%. We estimate the vertical resolution of the measurement by moving
the upper boundary of the grain packing in a partly filled tube through the sensor
and recording the output reading for each location: the corresponding capacitance
variation is roughly linear and spreads over 2 mm which therefore represents the
typical spatial resolution of the measurement.
A Spectral Dynamics SD195 signal analyser records instantaneous variations of
the readings of four sensors (up to a 200 kHz sampling frequency), while averages of
10 physical parameters over time intervals of 1 s are recorded simultaneously.
Another important issue is the determination of the various characteristic velocities
of the flow. Mean air and grain velocities can be estimated by combining grain and air
flow rate values with those of the particle fraction; on the other hand the velocity of
the plugs in the wave regime can be determined from spatio-temporal diagrams of the
particle density distribution along the pipe. Such diagrams were also constructed for
the free-fall regime and provide the velocity of small density fluctuations associated
with small aggregates of particles. Spatio-temporal diagrams are obtained by means
of a digital linear CCD camera mounted on a micrometric assembly allowing a
precise alignment of the view field of the camera with the flow tube. The readings
of all the pixels of the linear sensor are directly transmitted to a computer at a
repetition rate of 2000 Hz for the free-fall regime and 500 Hz for the wave regime.
Spatio-temporal diagrams obtained in this way are stored as numerical data arrays
which can be visualized by standard image processing programs; in addition they
allow one to detect with great sensitivity oscillations and intermittencies of the flow
and to measure the propagation velocity of the perturbations.
2.2. Flow regime dependence on granular flow rate and humidity

The key parameter determining the flow regime is the granular flow rate, but the
degree of humidity H of the ambient air also significantly influences the flow regimes
observed. Breathing into the pipe for a second generates enough humidity on the
walls to completely block the flow of grains in the tube. More generally, large values
of humidity often induce clogging due to capillary forces between particles or between
particles and tube walls. On the other hand, if the atmosphere is too dry, electrostatic
forces appear and may also lead to a blocking of the flow. Humidity is monitored
by an electronic hygrometer and adjusted by a humidifier in order to obtain a good
reproducibility during a series of experimental runs.
The free-fall regime characterized by high velocities and low densities of particles
is only observed for grain flow rates with q > 0.35 m s−1 with little or no constriction
at the bottom (figure 1b). When the constriction is increased, the flow switches
to a regime with density waves of constant velocity in the range of flow rates
0.11 < q < 0.26 m s−1 (and for relative humidities between 45% and 55%). In this case,
a system of high-density plugs and low-density bubbles moves at a constant downward
velocity (figure 1c). For narrower constrictions and q of order 0.085–0.11 m s−1 , the
waves may start to oscillate with a system of plugs and bubbles moving up and down
with an amplitude of the order of a centimetre: this is particularly the case at low
humidity. At still lower flow rates q < 0.08 m s−1 , a compact flow is observed with
a slow constant velocity or an intermittent motion of the particles. This regime, in
which solid contact forces play a crucial role, will be discussed in a future paper and
we shall deal here only with free-fall and wave regimes.
In order to demonstrate qualitatively the influence of the contact forces in the
wave regime, we report near the end of the paper (§ 5) a few experiments realized
with beads which had been processed many times through the experimental tube
and had their surface roughened by collisions with the walls and with each other.
Time (ms)
0 100 200
100
Dz (mm)
200
Figure 2. Spatio-temporal images of the fluctuations of particle distribution along the flow tube
obtained with a linear camera in the free-fall regime. The dashed line corresponds to a constant
velocity of 2.3 m s−1 . Line sampling rate: 2000 Hz, height of field of view: 256 mm, duration of
recording: 240 ms, ∆z is the distance from the top of the tube.
Otherwise, all experimental results presented here correspond to beads used in only a
few experiments and displaying a smooth surface.
3. Stationary free-fall regime

3.1. Experimental observations
In this regime, a fast flow of the particles is observed, with no apparent velocity
fluctuations (no detectable flow rate variation is observed by measuring the mass
of grains flowing out of the tube either). Pressure fluctuations measured on the
local transducers are also of very small amplitude (1–2% of the average value)
confirming the near stationary nature of the flow. The superficial velocity of the
grains (typically q = 0.35–0.55 m s−1 ) is large and associated with a large downward
air flow (qa = 2–3 m s−1 ). This regime can be compared to the flow of suspensions
(Leung & Jones 1978; Chen et al. 1984) down a standpipe: the superficial velocities
of the grains are similar (even though the pipe diameter is ten times smaller in our
work) but the particle fraction is estimated to be of the order of 60% (instead of
around 20% in our experiments). Since this free-fall flow regime is stationary and the
50 0
(a) (b)
40
–1000
Pressure (Pa)
30
c (%)
20 –2000
10
–3000
0
0 0.5 1.0 0 0.5 1.0
z (m) z (m)
Figure 3. (a) Particle fraction variations in a free-fall flow along a 1.25 m long tube as a function of
•
the distance z from the top of the tube for a grain flow rate q = 0.56 m s−1 . , experimental values.
Lines correspond to theoretical predictions under different assumptions: - - - - -, friction force on
particles given by the Stokes formula; – –, Friction force corrected for finite Reynolds number and
finite particle fraction effects and wall friction force Fw equal to zero; —–, Friction force corrected
for finite Reynolds number and particle fraction effects and wall friction force Fw equal to 15%
of grain weight. (b) Time-averaged relative pressures with respect to atmosphere at four different
•
distances z. , q = 0.56 m s−1 ; , q = 0.54 m s−1 (the atmospheric pressure pat is taken equal to
zero). Lines correspond to theoretical predictions for q = 0.56 m s−1 under the same asssumptions
in 3(a).
characteristic parameters vary slowly along the tube, it represents a suitable case to
test our measurements against a simple theoretical model of the flow.
Spatio-temporal diagrams of the density fluctuations along the tube (figure 2)
display parallel striations reflecting the motion of small grain clusters and/or of local
density fluctuations (the spatial resolution, typically 1 mm, is not good enough to
detect the motion of individual grains). The lines are slightly curved at the upper part
of the tube (reflecting the flow acceleration) and reach a constant slope (dashed line
in figure 2) at a typical distance of 0.5 m below the hopper: this slope corresponds to
a large constant velocity v∞ of order 2.3 m s−1 .
These qualitative observations can be compared to the particle fraction profiles
c(z) along the tube: the particle fraction decreases quickly (figure 3a) and reaches a
limiting value c∞ of order 0.16 at about 0.5 m below the top. This is accompanied by
an increase of the grain velocity v(z) with distance since the superficial velocities q
and qa satisfy (assuming a steady state):
q = c(z)v(z), (3)
qa = (1 − c(z))va (z). (4)

The low value of c∞ gives a large difference between the superficial velocity q
and the limit velocity v∞ . The increase of v(z) is in qualitative agreement with the
spatio-temporal diagram of figure 2. Equation (4) indicates that air velocity must, on
the other hand, decrease with distance. The flows of grain and air induce, in turn,
pressure variations displayed as a function of z in figure 3(b) for different velocities
0.45 < q < 0.55 m s−1 (the atmospheric pressure pat will be taken equal to zero in the
following). A large underpressure region (3000 Pa lower) appears in the upper part
of the tube, while in the region z > 0.5 m where c(z) has reached its limiting value,
pressure increases linearly until it reaches atmospheric pressure (pat = 0) at the outlet.
The point at z = 0 represents the pressure measured in the hopper above the grain
packing and is very close to zero: a strong pressure gradient above z = 0.2 m results,
which explains the large downward air flow through the porous grain packing.
Let us now write the equations of motion for the air and for the grains (Ginestra
et al. 1980; Chen et al. 1984): combined with the mass conservation equations above
they will allow us to determine friction forces on the walls from pressure and particle
fraction profiles without needing further relations. Assuming in addition a simple
expression from the literature for the friction force between air and the grains, it will
then be possible to predict the full pressure and particle fraction profiles along the
pipe for given flow rates q and qa .
3.2. Equations of motion for the air and for the grains and estimates of friction forces
It is assumed here that the flow is stationary in time and homogeneous across a
given section of the tube. Then, the velocities va and v of air and of the grains
respectively depend only on the distance z from the bottom of the hopper (taken
positively downwards) and are related to the particle fraction c(z) by equations (3)
and (4). Let us first assume that the friction force between the grains and the tube
walls is negligible. Then, when the flow is stationary, the Lagrangian acceleration
dv(z)/dt = v(z) dv(z)/dz satisfies
dv(z)
mv(z) = mg − F(v(z) − va (z)). (5)
dz
Here m is the mass of a single grain given by m = 4πρg a3 /3 and F(v(z) − va (z)) is
the friction force on the grain, which depends only on the relative velocity of the
particles with respect to the air. On the other hand, the equation of motion of air can
be reduced to
dp F(v(z) − va (z))
= ρg c(z) , (6)
dz m
in which ρg c(z)/m is equal to the number of particles per unit volume. Equation
(6) assumes a balance between the pressure gradient and the friction force between
the air and the grains: this amounts to neglecting both the Lagrangian acceleration
ρa v(z) dv(z)/dz and the friction force of air on the walls. The first term is negligible
due to the low density of air and the second is shown below to be markedly smaller
than the pressure gradient. Equation (6) may be combined with (5) to obtain the
following equation which does not include explicitly the friction force F:
dp dv(z)
= ρg gc(z) − ρg c(z)v(z) . (7)
dz dz
At large enough distances from the hopper (z > 0.4 m), the particle fraction c(z) and
the velocities v and va reach the limiting values c∞ = 0.16, v∞ and va∞ (figure 3), and
the term ρg c(z)v(z) dv(z)/dz is equal to zero in (7). However the value of (dp/dz)
predicted by taking c∞ = 0.16 in (7) is of the order of 3900 Pa m−1 which is larger
than the experimental value (dp/dz) = 3500 Pa m−1 at long distances (figure 3). This
difference implies that the friction force on the walls is not zero and that (7) must be
replaced (in the constant-pressure-gradient and velocity region) by
dp(z) Fw
= ρg gc∞ − . (8)
dz πR 2
Fw is the friction force on the walls per unit length of the tube and is of the order of
15% of the grain weight for q = 0.56 m s−1 (as in figure 3a). Let us show now that Fw
is made up mostly of the friction force Fwg between grains and the walls and that the
component Fwa due to air friction with the walls is negligible. Estimating Fwa is not
straightforward since the flow profile is strongly influenced by the grains (estimating
the frictional force between the grains and the walls would be even more difficult).
The typical Reynolds number Re of the global air flow in the tube defined as 2Rvm /ν
is of order 400 (for a mean air velocity vm of the order of the limiting velocity 3 m s−1 ).
An upper bound on the pressure drop component due to air friction with the walls
may be obtained by considering a characteristic roughness ε of the tube walls of the
order of the grain radius a. A classical (Olson 1970) approximate expression for the
characteristic friction velocity u∗ is given by the relation:
u
= 2.5 ln(r/ε) + 8.5,
u∗
where r is the radial position in the pipe. After integration, one obtains
vm
u∗ = .
2.5 ln(R/ε) + 7.25
With the values of the physical parameters given above, one obtains u∗ ≈ 0.2 m s−1
and vm ≈ 3 m s−1 , which corresponds to the following pressure gradient:

∂p 2
= f ρa vm , (9)
∂z D 2
where f is a friction factor which depends on the Reynolds number and the rugosity
and is of the order of 0.06. Thus, one finds |∂p/∂z| ≈ 110 Pa m−1 . This justifies the
assumption that this contribution to the pressure gradient term is negligible.
3.3. Computation of theoretical pressure and particle fraction profiles and comparison
with experimental results
In order to determine the profiles p(z) and c(z), one first needs to compute the friction
force F(v(z) − va (z)). At the relative velocities of the particles with respect to the
air found in our experiments (3 m s−1 at most), the corresponding Reynolds number
Rea = 2a(v − va )/ν (ν being the kinematic viscosity of air), is always below 100 at
low concentrations. Then, in the low particle concentration limit, the force F may be
approximated by the empirical relation (Olson 1970)
3
1/2
F(v − va ) = 6πηa(v − va ) 1 + 16 Rea (10a)
(η is the dynamic viscosity of air), which reduces at very low Reynolds numbers to
the classical Stokes relation. In our experimental case, the dilute approximation is not
valid since measured particle fractions range between 15% and 25% and the particle
fraction is still higher at the upper end of the tube. We take into account these finite
concentrations by introducing a correction factor estimated from Richardson & Zaki’s
(1954) classical relation between the sedimentation velocity vs (c) of a suspension and
the Stokes settling velocity for a single isolated particle vstokes . This equation is written
as vs (c) = vstokes (1 − c)n in which the exponent n is taken equal to 5.5 at low Reynolds
numbers: this implies that the coefficient 6πηa in equation (10a) must be multiplied
by (1 − c)n−1 (the additional factor 1/(1 − c) reflects the fact that sedimenting particles
induce a backflow of the suspending fluid so that the relative velocity is actually
vs (c)/(1 − c)). Equation (10a) then becomes
3
1/2 1
F(v − va ) = 6πηa(v − va ) 1 + 16
Rea . (10b)
(1 − c)n−1
Combining equations (3)–(5) and (10b), one obtains the following first-order differ-
ential equation for c(z) (assuming at this stage Fw = 0):
1/2
dc c3 6πηac3 q qa 3 a q qa 1
= −g 2 + − 1 + − , (11a)
dz q mq 2 c 1−c 8ν c 1−c (1 − c)n−1
which, for low Reynolds numbers and low particle fractions, becomes

dc c3 6πηa c2 c3 qa
= −g 2 + − 2 . (11b)
dz q m q q (1 − c)
Equations (11a) and (11b) are then integrated numerically to obtain the theoretical
particle fraction profiles c(z) displayed in figure 3(a). Note that the value c(0) right at
the top of the hopper cannot be determined precisely due to the fast variation of c(z)
in this region: on the other hand (and for the same reason) the choice of c(0) does
not markedly influence c(z) in regions of slower variation farther from the hopper.
Even a large variation of c(0) corresponds to an uncertainty of a few mm in the z
scale of the profiles, which has a small influence on the value of c(z), except very close
to the hopper. We have chosen the value c = 0.5, which is a good order-of-magnitude
estimate of the particle fraction just after decompaction, to obtain the curves of
figure 3(a). In order to determine c(z), we have used in (11a, b) the experimental value
of the air flow qa . This air flow might alternatively be estimated for a given grain flow
q from the permeability and particle fraction of the grain packing in the hopper and
the pressure variation across this packing. This estimate would however be much less
precise than the direct measurement in view of the uncertainty in these parameters.
One observes first in figure 3(a) that the fit of the theoretical curves with the
experimental data is improved by including finite Reynolds number and particle
fraction effects. The theoretical profiles also confirm the experimental observation
that c(z) reaches a lower limit c∞ at a distance of 0.4 m. This limiting value c∞ is equal
to 0.14 when the friction force Fw is omitted. In order to obtain the experimental
value 0.16, a friction force Fw equal to 15% of the grain weight must be introduced,
as already found in the previous section (this amounts to decreasing the gravity
acceleration g by 15%). The third curve in figure 3(a) corresponds to this hypothesis
and differs little near the hopper from that corresponding to Fw = 0, while the fit
becomes better further down the tube. Note that the value 0.15 of the ratio between
Fw and the grain weight can be determined directly from (7) and does not depend
on the expression for the friction force F between the air and the grains. This ratio
should depend however on various flow parameters such as the flow rate, the surface
condition of the grains and of the tube and their diameters. Similar observations have
been reported on pressure gradients in circulating fluidized beds (although at higher
particle fractions) by Srivastava et al. (1998) and were also interpreted in terms of
wall friction.
Theoretical pressure profiles p(z) are computed at the same time as c(z) by means
of (6) (at each step of the integration, the friction force F is computed by combining
equations (3), (4) and (11a, b)). The integration constant is then chosen to give the
atmospheric pressure at the lower outlet of the tube. Closer to the top of the tube,
the fit between experimental and theoretical pressure variations is markedly better
when the effects of the finite Reynolds number and particle fraction values are taken
into account (figure 3b); it is further improved by introducing the friction force Fw
of the grains with the walls.
5
3
v, va (m s–1)
0
0 0.5 1.0
z (m)
Figure 4. Air and grain velocity variations in a free-fall flow (q = 0.56 m s−1 ) along a 1.25 m long
tube as a function of distance z from the bottom of the hopper. Theoretical grain and air velocity
profile: v(z) (—) and va (z) (- - -). N, Grain cluster velocity determined from spatio-temporal diagram
•
at z = 0.4 m; , mean grain velocity computed from experimental particle fraction profile.
3.4. Particle and air velocities in the stationary free-fall regime

By combining the two mass conservation equations (3) and (4) and the particle
fraction profile c(z), one obtains the respective profiles va (z) and v(z) of the mean
air and particle velocities. Theoretical curves obtained by taking into account the
finite Reynolds number and density effects on the drag as well as grain friction with
the walls are plotted in figure 4. One observes a decrease of the air velocity with
distance (corresponding to the decreasing particle fraction) while, on the contrary,
the grain velocity increases. In the upper part of the tube, the air moves faster than
the grains and contributes to their acceleration. The velocities become equal around
z = 0.2 m where pressure is minimal. Then the grains move faster than the air and
their acceleration decreases until their weight is exactly balanced by friction with the
air and with the walls.
These data are first compared to semi-experimental values determined by substi-
tuting into equations (3) and (4) experimental particle fractions plotted in figure 3(a).
The good agreement between the two set of values essentially mirrors the agreement
between theoretical and experimental particle fraction variations. More meaning-
ful comparisons can be made with direct velocity determinations using the spatio-
temporal diagram of figure 2. As already stated, the particle density fluctuations give
rise to a set of parallel dark and light lines: the slope dz/dt of these lines is equal
to the propagation velocity of the fluctuations which is thus defined well at a given
height. The velocity determined in this way at z = 0.4 m has been plotted on figure 4.
It is close to the theoretical particle velocity although slightly lower. The difference
is probably due to the fact that the linear camera does not have a sufficiently good
spatial resolution to detect individual particles. Thus the fine dark and light lines of
figure 2 probably correspond to small clusters of particles or to small low-density
zones which may have a different velocity from that of individual grains: the fair
agreement displayed in figure 4 indicates that these diagrams however represent a
good practical way to track particle motions quantitatively.
Studying the stationary free-fall regime has allowed us to test measurement and
modelling tools in a simple experimental case. It has been possible in particular to
estimate quantitatively the magnitude of friction forces of the particles with the walls
without making assumptions about the expression for the friction force between the
air and the particles. Using a simple classical model for this latter force allowed us in
addition to predict precisely pressure and particle fraction variations with distance.
The estimate of particle cluster velocities from spatio-temporal diagrams has also
demonstrated the potential of this technique for analysing the flow dynamics.
We shall now apply the same approach to the analysis of the density-wave regime
characterized by a sequence of plugs and low-density bubbles moving downwards at
a constant velocity vw (Raafat et al. 1996). This system is more complex and new
characteristic variables must be introduced such as the velocity vw and the length of
the bubbles and the plugs. Both the particle fraction and the air and grain velocities
differ in the plug and bubble regions and the flow can be considered as stationary only
in the reference frame moving at velocity vw . In this part of the work, our objective is
to analyse the dependence of the structure of the flow (particle fraction distribution,
plug and bubble lengths, plug, particle and gas velocities, etc.) on the grain flow rate
and to determine the relations between the pressure gradient, the wall friction forces
and the various experimental variables.
4. Constant-velocity density-wave regime

4.1. Spatio-temporal characteristics of the constant-velocity wave regime
Spatio-temporal diagrams obtained in the wave regime at two different flow rates are
displayed in figure 5(a, b). The sequence of plugs and bubbles appears as inclined
stripes of slope corresponding to the plug velocity vw . This slope is quite constant for
a given plug and also from one plug to another so that vw can be determined precisely.
The variation of vw with the flow rate q is seen in figure 6 to be roughly linear for
large q(q > 0.16 m s−1 ) and constant for small q: plugs move faster near the transition
to the free-fall regime where they disintegrate. Extrapolating the variation of figure 6
suggests that vw would go to zero at a positive non-zero value of q (transient-wave
regimes with stationary plugs were also observed in a tube closed at the bottom and
filled from the top). A net downward grain flow from one plug to the next is observed.
Inside each bubble, the spatio-temporal diagram is qualitatively similar to that in the
free-fall regime: the curvature of the lines (figure 5a) indicates an acceleration of the
grain clusters as they fall out of a plug. The distribution of the plugs is variable: in
some cases plugs and bubbles are well-defined and separated while, in others, plugs
cluster together (but remain separated) and buildup a like kind of ‘puff pastry’ as can
be seen in figure 5(a).
The sequence of plugs and low-density bubbles does not extend down the full length
of the tube: in the top part, it becomes established only after a distance of 4 to 15 cm
below the hopper (increasing with flow rate q) where short plugs appear and quickly
grow to their final length (figure 5a, b). Closer to the hopper, the spatio-temporal
diagram is similar to that observed in the free-fall regime (figure 2) with a velocity
increasing with distance. At the other end, the lower part of the tube (not visible on
the figure) is occupied by a compact bottom plug of length lbp decreasing from more
than 300 mm down to 50 mm as q increases (figure 9).
Information provided by the spatio-temporal diagrams is complemented by particle
fraction measurements. Figure 7 displays the time variation of the capacitance sensor
(a) (b)
Figure 5. Spatio-temporal images of the fluctuations of particle distribution along the tube obtained
with a linear camera in the wave regime. Line sampling rate: 2000 Hz, height of field of view:
450 mm, time lapse corresponding to diagrams: 0.5 s. (a) q = 0.14 m s−1 , (b) q = 0.22 m s−1 . The
spatio-temporal diagrams extend from the top of the tube (the bottom of the hopper appears as a
dark strip at the top). Both horizontal white lines correspond to mechanical supports and the dark
stripe below to a pressure transducer.
0.8
0.6
vw (m s–1)
0.4
0.2
0
0.10 0.15 0.20 0.25
q (m s–1)
Figure 6. Wave velocity vw as a function of the superficial grain velocity q in the steady waves
•
regime (vw is computed from the spatio-temporal diagram). , Data points corresponding to beads
with a smooth surface; N, data points for beads with a rougher surface.
60
Particle fraction inside plugs
54.5%
50
40
c (%)
30
Average particle fraction
25%
20
10 8.5%
Minimum particle fraction inside bubbles
0
0 5 10 15
Time (s)
Figure 7. Time variation of the reading of a capacitive particle fraction sensor located at 300 mm
below the hopper in the density-wave regime (q = 0.1 m s−1 ), particle fraction inside plugs = 54.5%,
average particle fraction = 24%, minimum particle fraction inside bubbles = 8.5%.
in the wave regime, showing large fluctuations of the particle fraction between well-
defined high values of often 50% or more in the plugs and low values of 15% or
less inside bubbles. The asymmetrical peaks correspond to the plugs: the sharper rear
slope of the peak corresponds to the upper side of the plugs where the particle fraction
increases abruptly as grains falling through a bubble hit the top of the plug. On the
other hand, the particle fraction variation is smoother at the bottom of the plugs
where beads accelerate progressively as they leave it. The particle fraction cp in the
plugs is more or less constant from one plug to the next while the minimum particle
fraction cb inside bubbles increases with their length. This reflects the acceleration
of particles inside bubbles and the fact that they reach a higher velocity at the
bottom of long bubbles. The distribution of the size of the plugs and the bubbles
can be estimated from the width of the peaks and their intervals together with direct
measurements of the plug velocities: in a previous work (Raafat et al. 1996), the
size of the plugs was found to be relatively constant and of the order of 10 mm
while bubbles are generally larger and with more dispersed sizes (no indication of
very broad size distributions, such as power laws for instance, was found however).
Variations of the mean bubble sizes with the flow parameters are displayed below in
figure 9.
Figure 8 presents, for beads with a smooth surface, variations with the superficial
velocity q of four specific particle fractions: the fraction cp in the plugs, the minimum
fraction cb at the bottom of long bubbles, the time-averaged fraction c̄ and the
fraction cbp in the bottom plug near the outlet of the tube. The particle fraction cp
in the plugs decreases from 60% for q = 0.11 m s−1 (close to the particle fraction of
a fixed random packing) down to 30% for q = 0.23 m s−1 with an acceleration of the
variation close to q = 0.16 m s−1 (the same value above which the wave velocity vw
increases faster with q). The minimum fraction cb in the bubbles is of the same order
of magnitude as in the free-fall regime; the time-averaged fraction c̄ is of the order
of 25% at low flow rates and decreases quickly with q towards an almost constant
value of 20%. The particle fraction cbp in the bottom plug is larger than that of the
60
50
40
c (%)
30
20
10
0
0.10 0.15 0.20 0.25
q (m s–1)
•
Figure 8. Variations of time-averaged particle fraction c̄ (N), particle fraction in plugs cp ( ),
particle fraction inside bubbles cb (), and bottom plug particle fraction cbp (H) in the wave regime
as a function of superficial grain velocity q.
2
lpb, lbp (m)
0.1
0.10 0.15 0.20 0.25

q (m s–1)
•
Figure 9. Variation of the mean plug+bubble length lpb = lp + lb ( ) and of the bottom plug
length lbp (N) as a function of the superficial grain velocity q.
moving plugs and is approximately constant and of the order of 55% for q below
0.16 m s−1 (5–7% less than a fixed packing); it then decreases towards 45%.
Spatio-temporal diagrams and particle fraction variations with time also allow us
to determine the mean length lpb of the elementary cells including a plug and a bubble.
It is taken equal to the mean time interval between two particle fraction peaks divided
by the wave velocity vw : its variations and those of the length of the bottom plug
lbp are plotted in figure 9 as a function of q. The mean value of the cell length lpb
is almost constant with q and of order 40 mm up to the transition to the free-fall
regime. On the other hand, the length of the bottom plug steadily decreases: at the
transition between the wave and free-fall regimes, lbp is less than 50 mm.
qa, qabp, qaw , qar, qarbp (m s–1)

0.4
0.2
0.10 0.15 0.20 0.25

q (m s–1)
•
Figure 10. , Mean superficial velocity qa of air. e, Superficial air velocity component qabp
corresponding to passive entrainment in the bottom plug. H, Superficial air velocity component
qarbp corresponding to the relative motion of air with respect to the grains in the bottom plug. N,
Superficial air velocity component qaw corresponding to wave motion. , Superficial air velocity qar
measured in a reference frame moving at velocity vw . All velocities are plotted as a function of the
superficial grain velocity q.
At this stage, the structure of the density-wave flows and the corresponding particle
distributions have been determined as a function of the grain flow rate: in all cases, one
observes a constant-velocity motion of dense plugs with a particle fraction sometimes
barely lower than that of a static packing and decreasing at large grain flow rates.
We shall now investigate the kinematical properties of the system by evaluating the
flow components of the air and of the grains, both associated with the global motion
of the wave system and relative to this moving reference frame.
4.2. Global and relative flow components
The mean superficial air velocity qa in the tube section is computed by means of
equation (2) from the reading Qmeas of the air flow sensor at the inlet of the hopper.
Figure 10 displays the variations of qa (filled circles) as a function of q: the velocity qa
increases at first linearly with q but is always 3–10 times smaller than in the free-fall
regime. Both q and qa result from the combination of a global mean motion at the
wave velocity vw and a relative motion of both the air and the grains with respect
to the plug boundaries. An interesting feature of the reference frame moving at the
velocity vw is that plugs and bubbles appear fixed: the flow distribution is therefore
stationary which makes writing conservation equations easier. We also assume that
the plug and bubble lengths lp and lb are constant with time and from one cell to
another.
Let us estimate now the relative magnitude of the different flow components: we
denote as ua and u the respective air and grain velocities in the moving reference
frame and as va and v the corresponding values in the laboratory frame. Then
v = u + vw , va = ua + vw . (12a, b)
The relative velocities ua and u are constant with time at a fixed point in the moving
frame. In the same way, let us denote as qar and qr the superficial velocities of the air
and the grains in the moving frame: they are constant with both time and distance
(due to mass conservation) and satisfy
qar = ua (1 − c), qr = uc, (13a, b)
in which c is the local particle fraction. The differences between qar and qr and the
corresponding superficial velocities qa and q in the laboratory frame are the volume-
averaged superficial velocities qw = vw c̄ and qaw = vw (1 − c̄) which represent the flow
rates associated with the global wave motion. The relative superficial velocities in the
moving frame thus satisfy
qar = ua (1 − c) = qa − vw (1 − c̄), qr = uc = q − vw c̄ (14a, b)
Variations of these superficial velocities qar , qaw , qr , qw as function of the global grain
superficial velocity q are plotted in figures 10 (for the air) and 11(a) (for the grains).
In agreement with previous results (Raafat et al. 1996) the relative superficial velocity
of the grains qr varies little with the flow rate (particularly for q > 0.12 m s−1 ) and
does not vanish when the wave velocity goes to zero.
Grain and air velocities in the different parts of the flow can now be determined
by using equations (12a, b) and (13a, b) and the particle fractions from figure 8.
Figure 11(b) displays variations of u and ua determined in this way both in the plugs
(up and uap ) and at the end of the bubbles (ub and uab ). Since velocity is continuous
at the lower plug–bubble boundaries, up is also the grain relative velocity at the top
of the bubbles. Figure 11(b) shows therefore that this relative velocity u increases by
a factor of 3 to 10 between the top and the bottom of the bubbles (in agreement
with the curvatures of the streaks in the spatio-temporal diagram); u is however
smaller than in the free-fall regime since the acceleration distance is shorter (the
terminal velocity is not reached) and the downflow of air accelerating the beads is
negligible. The relative air velocity qar is either very small or negative: this implies that
the upward permeation of the air through the plugs induced by the mean pressure
gradient balances the passive drag by the downwards motion of the grains.
An interesting issue is the value in the moving reference frame of the initial relative
velocity up of grains leaving a plug. We suggest that the bottom layers of the plug
detach sequentially one after another due to the dilatancy which is required to deform
the plug. Let us call τ the time interval between the detachment of two successive
grain layers, each of typical thickness 2a. Then up = 2a/τ. Let us assume that, after
the detachment, the motion of the layer corresponds to a free fall, neglecting air
friction. Then, the spacing hR between two layers at the time when the second layer
becomes detached should be hR = 12 gτ2 . One expects the value of hR to be a few
microns – a small percentage of the grain diameter. The values of hR computed in
this way from the experimental values of up displayed in figure 11(b) range between
4 and 15 µm, in agreement with these expectations (the variation is explained by the
variable particle fraction of the plugs which will modify the locking between grains).
A similar decomposition of the global air flow rate qa can be performed in the
bottom plug (taking this time the flowing grains as the moving reference frame). A
first component of qa is the volume qabp of air per unit time and unit area dragged
by the flowing grains:
q
qabp = (1 − cbp ). (15)
cbp
This value qabp and the difference qarbp = qa − qabp are also plotted in figure 10 (open
circle and downward triangle symbols). The difference qarbp would be equal to qa if
0.25
(a)
q, qr, qw (m s–1) 0.20 q
0.15
0.10
0.05
0
0.10 0.15 0.20 0.25
q (m s–1)
1.5
(b)
up, uap, ub, uab (m s–1)
1.0
0.5
0.10 0.15 0.20 0.25

q (m s–1)
Figure 11. (a) Variation of grain superficial velocity qr (N) measured in the moving frame and of
component qw (∆) corresponding to the global wave motion, as a function of the total superficial
velocity q (also displayed as a continuous line as a reference). (b) Variation of local grain velocity
•
relative to reference frame of velocity vw for air (index a) and grains in plugs (uap ( e) and up ( ))
and at the bottom of bubbles (uab (∆)) and ub (N)) as a function of grain superficial velocity q.
the air had no relative motion with respect to the grains in the bottom plug. This
is indeed the case at low flow rates (q < 0.14 m s−1 ) for which the bottom plug is
long (lbp > 200 mm from figure 9) and compact (cbp > 55% from figure 8). At higher
flow rates both the particle fraction and lbp decrease so that the relative motion of
the air and the grains is no longer negligible: as can be seen in figure 10, the global
flow rate qa is then markedly higher than qabp due to the downwards air flow induced
by the pressure gradient through the bottom plug (these permeation components are
estimated below).
The previous discussion has demonstrated that the density-wave flows can be
separated into a component representing passive transport of the air and the grains
at the wave velocity vw and components corresponding to their relative motion with
4000
3000
P (Pa)
2000
1000
0
0 250 500 1000
z (mm)
Figure 12. Pressure variations as a function of distance along the tube for different superficial
• ∗
grain velocities: N, q = 0.2 m s−1 ; , q = 0.19 m s−1 ; , q = 0.16 m s−1 ; ⊗, q = 0.35 m s−1 ; ,
q = 0.135 m s−1 . Dotted lines correspond to linear regressions over the data and to assumed
variations in the bottom part of the tube.
respect to the plugs–bubbles sequence. The relative air and grain flow rates in this
moving frame are almost constant with q as well as the mean length of the plug–
bubble pairs: this leads to the approximate picture of a flow with roughly constant
characteristics in the moving frame but dragged along at the velocity vw varying with
q. In order to obtain a more precise picture the pressure gradient distribution will
now be investigated: it plays a key role both in determining the permeation of the air
through the dense parts and in the balance of the weight of the grains so that they
reach a constant velocity.
4.3. Pressure profiles and pressure gradient in the density-wave regime
Time-averaged pressures obtained from the four sensors are displayed in figure 12
as a function of distance from the hopper, for different grain superficial velocities q.
In contrast with the case of the free-fall regime (figure 3b), the pressure inside the
tube is generally higher than the atmospheric pressure pat ; even at the very top of
the tube, it is of the order of pat . The pressure increases approximately linearly with
distance down the part of the pipe occupied by the sequence of plugs and bubbles.
Inside the bottom plug, the flow is steady: the particle fraction and air and grain
velocities can then be assumed to be constant with both distance and time so that the
pressure gradient is also constant with distance. Pressure decreases linearly, therefore,
along the bottom plug down to atmospheric pressure at the outlet (see dashed lines).
A data point fully compatible with this hypothesis is obtained for q = 0.35 m s−1 on
the pressure port located at z = 950 mm (figure 12) which at this flow rate is located
inside the bottom plug (for other flow rates, all pressure ports are inside the wave
region).
Finally, one observes that the pressure drop across the packed grains in the hopper
is small compared to that across the bottom plug (the pressure at the top of the tube
is close to atmospheric pressure). This is due to the fact that, in the hopper, pressure
gradients are localized over a distance of a few mm from the outlet due to its conical
shape while the bottom plug is several cm long and of constant, small, diameter. The
pressure profile measurements demonstrate therefore the important contribution of
15 000
10 000
∂ p (Pa m–1)
∂z
5000
0
0.10 0.15 0.20 0.25
q (m s–1)
Figure 13. Variations of pressure gradients in the wave regime as a function of the superficial grain
•
velocity q. Pressure gradients measured experimentally: N, for rough grains; , for smooth grains.
Pressure gradients corresponding to the weight of the grains: e, for smooth and 4, rougher grains.
the bottom plug to the stability of the density-wave regime: it allows the build-up of
a strong pressure gradient along the plug–bubble sequence while avoiding inducing a
large downward air flow which might destroy the plugs (as happens near the transition
between the wave and free-fall regimes).
Figure 13 (filled circles) displays variations with the velocity q of the experimental
pressure gradient ∂p/∂z (equal to the slope of the linear part of the curves of
figure 12): ∂p/∂z decreases by about 30% as q increases from 0.1 to 0.23 m s−1 .
In the two next sections, we analyse how these pressure data and momentum
conservation equations can be used to determine first the wall friction in the wave
regime and then permeation flows due to the non-zero permeability of the dense
regions (moving plugs and bottom plug) which play a very important part at high
flow rates in the transition towards the free-fall regime.
4.4. Momentum conservation and wall friction in the density-wave regime

In order to determine the wall friction force, we shall use a similar approach as in
the free-fall regime and write the equivalent of equation (8) as
∂p (Fw )bp
= ρg gc̄ − (16a)
∂z (lp + lb )πR 2
Here, ∂p/∂z is the pressure gradient averaged over the bubble–plug cell. Equation
(16a) is written in the reference frame moving at the wave velocity vw so that the flow
can be considered as stationary: it reflects the balance among the pressure gradient,
the mass of the grains and the friction force (Fw )bp over the length of a bubble + plug
cell. This equation has been written over a cell length in order to be able to consider
than the momentum of the particle entering the cell is equal to that of particles
leaving it. An equivalent relation can be written over a sequence of consecutive cells
leading to the more general equation
∂p F̄w
= ρg gc̄ − . (16b)
∂z πR 2
Here, F̄w is the mean friction force per unit length – taken positive when oriented
towards the bottom. Equation (16b) will remain approximately valid even if the
sequence is not exactly periodical provided the momentum difference between particles
leaving and entering at the end of the sequence is not too large. As in the free-fall
regime, and since the air flow velocity is much lower than in this latter case, we shall
consider that the friction force of the air with the walls is negligible and that F̄w is
mostly due to the friction force of grains with the walls.
In order to estimate F̄w we have plotted on figure 13 (open circles for smooth new
beads) variations of ρg gc̄ computed from the data of figure 8. At a given q, the weight
ρg gc̄ is barely lower than ∂p/∂z so that, taking into account the experimental errors,
Fw should not be higher than 10% of the grain weight, i.e. certainly lower than in
the free-fall regime. We conclude that, for the corresponding set of smooth beads, the
momentum balance largely results from an almost ‘hydrostatic’ equilibrium between
the weight of the grains and the pressure gradient. Since most of the mass of the
grains is localized in the plugs, we expect the pressure gradient also to be higher inside
the plugs than inside the bubbles. This low wall friction and the weak dependence
of the pressure gradient on the grain flow rate confirm the previous intuitive picture
already suggested at the end of § 4.2, namely that the flow in the moving reference
frame depends very little on the global grain flow rate q, which affects only the global
drift of the plug–bubble sequence.
4.5. Permeation of air through the moving plugs
The permeation of air through the plugs is a key factor limiting the stability of the
density-wave regime: it is driven by the pressure drop across the plugs (which is
large enough to balance most of their weight) and limited by their permeability. If
the particle fraction in the plugs decreases with q (as in figure 8) their permeability
increases: the flow rate necessary to build up a pressure gradient balancing the weight
of the grains becomes too large and the transition towards the free-fall regime occurs.
The relative air–grain velocity of the permeation flow is related to the pressure drop
∆pp over the length of a plug by Darcy’s law:
K(cp ) ∆pp
(uap − up )(1 − cp ) = − (17)
η lp
in which K(cp ) is the permeability of the plug, η is the viscosity of air and the
factor (1 − cp ) is used to determine the superficial flow rate per unit area. As a first
approximation, one can estimate that ∆pp , is of the order of the pressure drop lpb ∂p/∂z
over the full cell (the friction forces should be concentrated in the plugs and most
of the grain mass which controls the term ρg gc(z) is also localized there). Table 1
lists ∆pp values obtained in this way using data from figures 9 and 13. Estimating
lp from the spatio-temporal diagrams then allows determination of an experimental
plug permeability value K(cp ) by applying (17) and using values of cp , up and uap
displayed above in figures 8 and 11(b). The corresponding values of K(cp ) and lp
are also in table 1 together with a theoretical estimate KCK of the plug permeability
obtained from the classical Carman–Kozeny relation:
(1 − c)3 d2
KCK = , (18)
180 c2
valid for packings of monodisperse spheres of particle fraction c and diameter d (d is
equal to 175 µm and c is taken to be cp ). The two values are in fair agreement, taking
into account the uncertainty in the values of the plug length lp and of the pressure
q (m s−1 ) ∆pp (Pa) lp (m) K (Darcy) KCK (Darcy)

0.104 295 0.01 25 30
0.107 260 0.01 39.5 37.5
0.111 215 0.01 60 63
0.121 213 0.01 33.5 77
0.132 168 0.01 90 85
0.140 181 0.01 38 104
0.163 208 0.008 85 140
Table 1. Variations with the grain superficial velocity in the density-waves regime of the pressure
drop on the plugs, the plug length lp and the plug permeability as estimated from the experimental
results (K) and from Carman–Kozeny’s formula using the measured plug porosity (KCK ).
drop ∆pp over the plug. No values are listed for larger flow rates (q > 0.16 m s−1 ): the
uncertainties are indeed too large due to the increase of the pressure gradient in the
bubbles compared to that in the plug and to the decrease of lp which becomes very
difficult to determine precisely. An important feature of the results of table 1 is the
strong increase (by a factor of 4) of the permeability with the grain flow rate q (this
increase should be even greater at higher flow rates): this confirms that the increase
of the permeation of air through the plugs at high flow rates will result in a transition
towards the free-fall regime.
4.6. Permeation of air through the bottom plug
The bottom plug observed in the present experiments also plays an important role.
First, it must withstand a pressure difference equal and opposite to that built up
along the plug–bubble sequence. The corresponding force is, in this case, oriented
in the same direction as the weight of the grains. The downwards resultant force
must then be balanced by friction forces on the walls and on the constriction at the
bottom of the tube (the complex geometry of the latter makes a direct evaluation
difficult). But the influence of the constriction should be felt only over a vertical
distance of the order of the Janssen length. Second, the finite permeability of the
bottom plug reduces the downflow of air induced by this pressure difference: if this
permeability is too large and/or the bottom plug is too short, the velocity of air
(and the wave propagation velocity) increases and the plugs–bubbles system finally
disintegrates. The relative flow rate qarbp of air through the moving grains due to the
finite permeability of the bottom plug is equal to qa − qabp . The experimental value
qarbp computed in this way is plotted in figure 10 and listed in table 2 as a function
th
of the grain superficial velocity q. As for the moving plugs, a theoretical value qarbp
of this relative flow rate can be obtained by estimating the permeability K(cbp ) of the
bottom plug through the Carman–Kozeny relation (18) using the measured value cbp
th
of the particle fraction; qarbp is then obtained by applying Darcy’s law to the bottom
plug:

th K(cbp ) ∂p L − lbp
qarbp = . (19)
η ∂z lbp
Here the pressure difference has been taken equal and opposite to the global pressure
difference along the total length (L − lbp ) of the wave section and ∂p/∂z is the
th
pressure gradient in this region. Both the values of qarbp and those of the permeability
K(cbp ) are listed in table 2. Both sets of values are of the same order of magnitude
up to q = 0.18 m s−1 : deviations seem due to experimental uncertainties. The larger
q (m s−1 ) qarbp (m s−1 ) Kbp (Darcy) th

qarbp (m s−1 )
0.104 0.015 27 0.006
0.107 0.032 27 0.007
0.111 0.027 33.5 0.013
0.122 0.022 37.5 0.022
0.132 0.038 44 0.047
0.140 0.122 44 0.055
0.163 0.142 46 0.091
0.167 0.148 57 0.105
0.185 0.145 85 0.206
0.206 0.168 140 0.409
0.223 0.188 140 0.464
Table 2. Variations with the grain superficial velocity of relative air superficial velocity with respect
to the moving grains in the bottom plug; qarbp = relative velocity determined from experimental flow
measurements, Kbp = permeability of bottom plug computed from experimental particle fraction
th
measurements by means of the Carman–Kozeny relation, qarbp = relative air velocity estimated from
the computed permeability of the bottom plug.
difference at higher flow rates may be due to spatial variations of the particle fraction
in the bottom plug near the transition between the wave and the free-fall regimes.
As observed above, the relative air flow in the bottom plug qarbp is almost negligible
below q < 0.14 m s−1 and increases at high flow rates due to the lower values of both
the bottom plug length and particle fraction. This increase of qarbp coincides with a
change in the behaviour of the other parameters, in particular a rapid increase of
the wave velocity vw and also a greater variation of the particle fraction in plugs. At
higher flow rates, the relative air velocity and the wave velocity increase more rapidly,
leading finally to a disintegration of the plug–bubble structure and to a transition
towards the free-fall regime.
This set of results provides a global picture of the mechanical equilibrium in the
wave regime for the type of grains we have used. The weight of the compact plugs is
mostly balanced by the air pressure gradient (almost as in an ‘hydrostatic’ equilibrium)
provided the permeability of the plugs is low enough to slow down their downwards
motion and also that the permeability of the bottom plug is low enough to reduce
the relative air flow through it. The friction of air and grains with the walls does not
seem to play a major role for the smooth beads which we have used. In the next
section, we investigate further these effects by performing the same experiments with
beads with a rougher surface, which should enhance wall friction.
5. Density-wave regime dependence on the surface characteristics of beads

All experimental results presented up to now have been obtained with beads used
only for a small number of experiments and which display a smooth and clean surface
when observed with a microscope. After many experiments have been performed with
a given set of beads, namely after they have passed a hundred times or so through the
tube, their surface becomes rougher and microscopical impurities attach to the surface.
Figure 14 displays particle fraction variations with the grain flow rate observed with
such a set of rougher grains. Comparing these variations with those displayed in
figure 8 for smooth beads, one observes that the particle fraction is always higher,
particularly at higher superficial grain velocities for which the mean particle fraction
60
50
c (%) 40
30
20
10
0
0.10 0.15 0.20 0.25
q (m s–1)
•
Figure 14. Variations of the time-averaged particle fraction c̄(N), of particle fraction in plugs cp ( )
and of minimum particle fraction cb () inside bubbles for rough grains in the wave regime as a
function of the superficial grain velocity q.
decreases much slower than in the previous case. The particle fraction in plugs is
also higher and is very close at low flow rates to the typical values (c = 63%) for
static bead packings. It can also be seen in figure 6 that the wave velocity vw is lower
than for smooth beads above q = 0.14 m s−1 and increases more linearly with q: the
change of regime above 0.15 observed for the smoother beads does not appear. These
results are confirmed by pressure gradient measurements displayed in figure 13. The
difference between the measured pressure gradient and the term reflecting the weight
of the beads is much higher than for smooth beads and represents up to 30–40% of
the total gradient: this implies that the friction force Fw with the walls is increased in
the same proportion.
We conclude that increasing the roughness of the bead surface enhances, as ex-
pected, the interaction between the grains and the walls; it also slows down signif-
icantly the wave motion and increases the particle fraction in the plugs and in the
mean.
6. Oscillating-wave regime
6.1. Experimental observations
At low grain superficial velocities near the transition towards the compact regime
(typically 0.085 to 0.11 m s−1 ), the velocity of the plugs and of the bubbles is not
always constant but in some cases oscillates around a non-zero drift velocity. Such
effects are particularly observed for grains with a rough surface used for a long time
in experiments. Figure 15(a) displays pressure variations with time measured at a
distance of 450 mm for q = 0.087 m s−1 . The global amplitude of the oscillations is
of the order of 10 000 Pa, which is very large and even higher than the mean value.
The frequency spectrum of the pressure is displayed in figure 15(b): the first peak
represents the oscillation frequency of order 5.7 Hz which is almost independent of
the grain flow rate.
Physically, these oscillations may be modelled as vibrations of a sequence of masses
(representing the plugs) connected by elastic springs (representing the bubbles): their
low frequency results from the large density of the compact plugs and the high
(a)
Pressure (Pa)
0
–7200
0 0.5 1.0 1.5 2.0

Time (s)
(b)
0
log (P( f ))
–5
–10
0 5 10 15 20
Frequency (Hz)
Figure 15. (a) Time variations of pressure in an oscillating density-wave flow (q = 0.087 m s−1 ) at
a distance of 450 mm below the hopper. (b) Logarithm of power spectrum of pressure variations in
the same experiment.
compressibility of the low-particle-fraction bubbles. An alternative point of view is to

consider the system as a two-phase flow (similar to air–water for instance) in which
both the mean density and the compressibility are high: they are characterized by
a very small sound velocity of the order of a few m s−1 resulting in low resonance
frequencies. There are two main damping mechanisms: wall friction, mostly in the
region of the plugs, and viscous dissipation, due to the flow of air through the plugs.
6.2. Simplified set of equations describing the oscillating-wave regime

We model these oscillations by adding to the previous equations time-dependent
terms corresponding to variations of the length lb of the bubbles, the plug velocity
vp and pressure (the plug length and particle fraction are taken constant). These
variables vary from one plug + bubble cell (denoted by an index j) to the next and
are assumed to oscillate about a mean value corresponding to the constant-velocity
regime discussed above (for instance the wave propagation velocity is still vw ). Each
parameter is then the sum of a constant part (no index) and of a periodic oscillatory
component (index 1), assumed to be small enough to neglect second-order terms. All
conservation equations are written in the reference frame moving at velocity vw . The
volume conservation of grains and of air between sections of two successive plugs is
then
1
dlbj
(cb (lb ) − cp ) = qj1 − qj+1
1
. (20)
dt
and
1
dlbj 1 1
dp1j
(cp − cb (lb )) = qaj − qaj+1 − χlb (1 − cb ) (21)
dt dt
(the superficial velocities q and qa are defined inside the plugs). The second equation
includes a term equal and opposite to the variation of the volume of grains and a
term due to the compressibility: χ = −1/v∂v/∂p of air. The particle fraction cb (lb ) is
the particle fraction at the bottom of the bubbles (the minimum value) and cb is the
average particle fraction in the bubbles.
The non-stationary part of the equation of motion of the plugs can be written as
dqj1 1
Fwj
ρ g lp = p1j−1 − p1j − , (22)
dt πR 2
1
where Fwj is the varying part of the friction force (assumed localized on the plug) in
cell j. A final equation relates, through Darcy’s law, time-varying pressure differences
across a plug to the relative velocity of air and grains inside it:
1 1 − cp 1 K(cp ) p1j−1 − p1j

qaj − qj = . (23)
cp η lp
Let us consider a progressive sine-wave perturbation of angular frequency ω and
wave vector k: the length of bubble j (the index starts at 1 for the topmost plug and
the bubble just below) and the pressure inside it satisfy
1
lbj (t) = lb1 exp[i(ωt − jk(lp + lb ))], (24)
p1j (t) = p1 exp[i(ωt − jk(lp + lb ))]. (25)

The time-varying components qj1 (t) and qaj
1
(t)of the air and gas flow rates have similar
expressions with complex amplitudes q and qa1 . Substituting these expressions into
1
(20–24) and neglecting the friction term Fw1 yields a system of linear equations which
has a solution if
4iωK(cp ) sin2 ( 12 k(lp + lb )) 4 sin2 ( 12 k(lp + lb ))
−ω 2 + + = 0. (26)
ηχlp lb (1 − cb ) ρg χlp lb cp (1 − cb )
This dispersion equation includes the damping effect of viscous dissipation of gas
flowing through the plugs (middle term). If this term is neglected, (26) closely resembles
classical dispersion relations for mass–spring chains. In the limit of long wavelengths
(low wave vectors such that k(lp + lb ) 1), ω/k tends towards a constant limit
representing the propagation velocity cosc of low-frequency oscillations of the plugs:
ω lp + lb
cosc = =p . (27)
k ρg χlp lb cp (1 − cb )
√
This velocity is of the order of the sound velocity 1/ χρg corresponding to an
hypothetical fluid with the compressibility χ of the air and the density ρg of glass.
Other factors are geometrical terms taking into account the relative volumes occupied
by air and grains.
6.3. Experimental orders of magnitude of the wave oscillation parameters

Using experimental values from previous sections (lb = 0.02 m, lp = 0.01 m), equation
(27) provides the estimate cosc = 23 m s−1 . The experimental oscillation frequency
fosc = 5.7 Hz corresponds therefore to a wavelength λosc = cosc /fosc = 4.1 m. The
oscillations observed can be considered as stationary resonant modes excited by
velocity fluctations at the outlet of the hopper: the hopper outlet (which communicates
with the outside atmospheric pressure through the low-hydraulic-impedance grain
packing inside the hopper) should be a pressure node. On the other hand, one expects
a maximum of the pressure oscillations at the bottom plug (there the displacement
amplitudes should be minimal). The lowest-frequency resonance mode should then
correspond to a length of the plugs–bubbles sequence equal to λosc /4 = 1 m. The actual
length is of order 0.8 m, so the agreement is reasonable in view of the crudeness of
the approximations involved.
In the low-frequency limit, the imaginary term in (26) which is partly responsible for
the damping of the oscillations is found equal to iω 3 K(cp )ρg cp /η. It may be compared
to the two other terms which are of the order of ω 2 . The experimental values yield a
ratio of the order of 0.1. This term therefore contributes significantly to the damping
of the oscillations.
A second dissipative mechanism is wall friction. The ratio of the wall friction term
1
Fwj /πR 2 in equation (22) and the acceleration term ρg lp dqj1 /dt is of order
1 1
Fwj g Fwj g
= :
πρg cp R lp g ωu
2 1 πρg cp R lp g ω δz 1
2 2
the first factor corresponds to the ratio between the oscillating component of the
friction force and the weight of the plugs, δz 1 is the typical amplitude of the plug
oscillations and u1 the corresponding velocity. For an amplitude of a few centimetres,
the second factor is of order unity so that the ratio of the fluctuations of the friction
force to the weight of the plugs characterizes the attenuation. In some cases, the
instantaneous friction force (constant + oscillating components) is large enough to
induce a complete blockage of the flow (Aider et al. 1999). A periodic stick-slip-like
motion of the wave system is then observed and can be considered as a high-amplitude
limit of the oscillating-wave regime.
7. Conclusion
In this paper, we have shown that the complex nature of the various granular
flow regimes in a vertical tube can be analysed in detail experimentally by combining
global measurements (air and grain flow rates, mean pressure gradients and average
particle fractions) with local information such as spatio-temporal diagrams and local
particle fraction variations. Combined with simple mass and momentum conservation
equations, these measurements enable one to determine important parameters, difficult
to measure directly experimentally, such as the wall friction or the permeation of air
through dense regions. The experiments have been realized in a tube of small diameter
which allows one to observe conveniently a variety of flow regimes. The same approach
might however be applied to other flow geometries and flow regimes.
In the stationary free-fall regime, particles accelerate under gravity until their weight
is balanced by friction forces with the air and with the walls; at the same time, the
particle fraction decreases with distance to ensure mass conservation. Friction on
the walls can be determined without further assumptions by comparing the particle
fraction and the pressure gradient at positions where the terminal velocity has been
reached (the wall friction is found to be of the order of 15% of the grain weight in
our experimental set-up). The full pressure and particle fraction profiles and the flow
distribution in the various parts of the system can then be determined quantitatively
at given air and grain flow rates: this requires only this wall friction value and a
simple model without adjustable parameters for the friction between air and the
particles.
In the steady wave regime, the upper part of the flow is occupied by a sequence
of plugs and bubbles moving down at a constant velocity: the weight of the plugs is
mostly balanced by an adverse pression gradient and, for a smaller part, by friction
of the grains on the tube walls. These friction forces represent only 10% of the weight
of the grains for beads with a clean, smooth surface and up to 20–30% for rougher
ones. In contrast with the free-fall case, pressure is higher inside than outside the tube
and the bottom part of the tube is occupied by a long, compact plug where pressure
decreases linearly towards atmospheric pressure at the outlet. Both the permeability
of the bottom plug and that of the moving ones must be sufficiently low to limit the
downflow of air which, if it is too large, accelerates the motion of the waves and leads
to their disintegration. Overall, and particularly when wall friction is low, density
waves appear as the superposition of a global drift of the plug–bubble sequence
and of a relative flow of grains with respect to the plugs of characteristics nearly
independent of the global flow.
Combining measurements of the grain and air flow rates with spatio-temporal
diagrams and particle fraction and pressure profiles allowed an estimation of particle
and gas velocities in the various regions of the flow. Other parameters difficult to
measure directly, such as the permeability of the moving or bottom plugs and the wall
friction, could also be determined by using these measurements as inputs for simple
mass and momentum conservation equations. In simple cases, the full pressure and
particle fraction profiles could be determined through simple additional hypotheses.
A similar approach was used in a simple case of unsteady flow (the oscillating-wave
regime) to predict the characteristic oscillation frequency.
Several problems remain to be solved, however, before a full understanding of
these flows is achieved. First, a theory predicting quantitatively the lengths of the
plugs and the bubbles still needs to be established (the size of the moving plugs
seems to be related mainly to the tube diameter). Second, we assumed throughout
the paper that the concentration and the velocity of particles is uniform across the
flow section. Significant three-dimensional effects are however to be expected in some
cases: for instance grains might tend to concentrate in the centre of the tube inside
the bubble regions which would induce a recirculation in the air flow. Indications of
such effects have been observed on close-up video recordings of flows in the density-
wave regime: grains seemed to concentrate on one side of the tube section inside
the ‘bubbles’, which might be due to a residual tilt of the tube. Recent theoretical
(Sinclair & Jackson 1989), Ocone, Sundaresan & Jackson 1993) and experimental
(Srivastava et al. 1998) studies have analysed these segregation effects in vertical
and inclined pipes: they predict significant variations of the particle concentration in
the radial direction, even for vertical tubes. However, more information, particularly
at the local scale, would be needed before attempting a quantitative comparison
with these models. A third issue is the dependence of the flow regime on the grain
shape and size and on the characteristics of their surface. Some preliminary re-
sults reported in § 5 have demonstrated the relevance of these properties. Finally,
other flow regimes such as an intermittent compact flow also provide challenging
questions.
We thank G. Chauvin, C. Saurine, O. Brouard and D. Vallet for their help in the
realization of the experimental set-up and of the capacitance sensors, N. Sommier
for her valuable contribution to the development of the capacitive particle fraction
measurement and of the spatio-temporal diagrams techniques and V. Terminassian
for his participation in the early phase of the experiments.
REFERENCES
Aider, J. L., Sommier, N., Raafat, T. & Hulin, J.-P. 1999 Experimental study of a granular flow
in a vertical pipe: a spatio-temporal analysis. Phys. Rev. E. 59, 778–786.
Behringer, R. P. & Jenkins, J. T. (Eds.) 1997 Proc. 3rd Intl Conf. on Powders and Grains. A. A.
Balkema.
Chen, Y. M., Rangachari, S. & Jackson, R. 1984 A theoretical and experimental investigation of
flow in a vertical standpipe. Ind. Engng Chem. Fundam. 23, 354–370.
Duran, J. 1997 Sables, Poudres et Grains. Editions Eyrolles Sciences.
Ginestra, J. C., Rangachari, S. & Jackson, R. 1980 A one-dimensional theory of flow in a vertical
standpipe. Powder Techn. 27, 69–84.
Jackson, R. 2000 The Dynamics of Fluidized Particles, Chap. 7. Cambridge University Press.
Jaeger, H. M. & Nagel, S. R. 1992 Physics of the granular state. Science 255, 1523–1531.
Lee, J. 1994 Density-waves in the flows of granular media. Phys. Rev. 49, 281–298.
Lee, J. & Leibig, M. 1994 Density-waves in granular flow: a kinetic wave approach. J. Phys. I
(Paris) 4, 507–514.
Leibig, M. 1994 Pattern-formation characteristics of interacting kinematic waves. Phys. Rev. E 49,
184–191.
Le Pennec, T., Ammi, M., Messager, J.-C., Bideau, D. & Garnier, J. 1995 Effect of gravity on
mass flow rate in an hour glass. Powder Technol. 85, 279–281.
Le Pennec, T., Måløy, K. J., Flekkøy, E. G., Messager J.-C. & Ammi, M. 1998 Silo hiccups:
dynamic effects of dilatancy in granular flow. Phys. Fluids 10, 3072–3079.
Le Pennec, T., Måløy, K. J., Hansen, A., Ammi, M., Bideau, D. & Wu, X. L. 1996 Ticking hour
glasses: experimental analysis of intermittent flow. Phys. Rev. E 53, 2257–2264.
Leung, L. S. & Jones, P. J. 1978 Flow of gas–solid mixtures in standpipes, a review. Powder Technol.
20, 145–160.
Lighthill, M. J. & Whitham, G. B. 1955 On kinematic waves: I. Flood movement in long rivers,
II. A theory of traffic flow on long crowded roads. Proc. R. Soc. Lond. A 229, 281–316 and
317–345.
Mountziaris, T. J. & R. Jackson, R. 1990 The effects of aeration on the gravity flow of particles
and gas in vertical standpipes. Chem. Engng Sci. 46, 381–407.
Musha, T. & Higuichi, H. 1976 The (1/f) fluctuation of a traffic current on an express way, Japan
J. Appl. Phys. 15, 1271.
Nott, P. & Jackson, R. 1992 Frictional-collisional equations of motion for granular materials and
their application to flow in aerated chutes, J. Fluid Mech. 241, 125–144.
Ocone, R., Sundaresan, S. & Jackson, R. 1993 Gas–particle flow in a duct of arbitrary inclination
with particle–particle interactions. AIChE J. 39, 1261–1271.
Olson, R. M. 1970 Essential of Engineering Fluid Mechanics, 2nd edn. London: International
Textbook Co.
Peng, G. & Herrmann, H. J. 1994 Density-waves of granular flow in a pipe using lattice-gas
automata. Phys. Rev. E 49, R1796–R1799.
Pöschel, T. 1994 Recurrent clogging and density-waves in granular material flowing through a
narrow pipe. J. Phys. I (Paris) 4, 499–506.
Raafat, T., Hulin, J.-P. & Herrmann, H. 1996 Density-waves in dry granular media falling through
a vertical pipe. Phys. Rev. E 53, 4345–4350.
Reydellet G., Rioual, F. & Clement, E. 2000 Granular hydrodynamics and density-wave regimes
in a vertical chute experiment. Europhys Lett. 51, 27–33.
Richardson, S. F. & Zaki, W. N. 1954 Sedimentation and fluidisation: Part I. Trans. Inst. Chem.
Engrs 32, 33–53.
Savage, S. B. 1979 Gravity flow of cohesionless granular materials in chutes and channels, J. Fluid
Mech. 92, 53–96.
Sinclair J. L. & Jackson R. 1989 Gas–particle flow in a vertical pipe with particle–particle
interactions. AIChE J. 35, 1472–1486.
Srivastava, A., Agrawal, K., Sundaresan, S., Karri, S. B. R. & Knowlton, T. M. 1998 Dynamics
of gas–particle flow in circulating fluidized beds. Powder Technol. 100, 173–182.
Wang C. H., Jackson R. & Sundaresan S. 1997 Instabilities of fully developed rapid flow of a
granular material in a channel. J. Fluid Mech. 342, 179–197.

Powder Flow Down A Vertical Pipe: The Effect of Air Flow

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Powder Flow Down A Vertical Pipe: The Effect of Air Flow

Uploaded by

Copyright:

Available Formats

c 2002 Cambridge University Press

J. Fluid Mech. (2002), vol. 459, pp. 317–345. 317

Powder flow down a vertical pipe:

(Received 2 March 2001 and in revised form 16 November 2001)

2.2. Flow regime dependence on granular flow rate and humidity

3. Stationary free-fall regime

qa = (1 − c(z))va (z). (4)

3.4. Particle and air velocities in the stationary free-fall regime

4. Constant-velocity density-wave regime

0.10 0.15 0.20 0.25

qa, qabp, qaw , qar, qarbp (m s–1)

0.10 0.15 0.20 0.25

q, qr, qw (m s–1) 0.20 q

0.10 0.15 0.20 0.25

4.4. Momentum conservation and wall friction in the density-wave regime

q (m s−1 ) ∆pp (Pa) lp (m) K (Darcy) KCK (Darcy)

q (m s−1 ) qarbp (m s−1 ) Kbp (Darcy) th

5. Density-wave regime dependence on the surface characteristics of beads

0 0.5 1.0 1.5 2.0

compressibility of the low-particle-fraction bubbles. An alternative point of view is to

6.2. Simplified set of equations describing the oscillating-wave regime

1 1 − cp 1 K(cp ) p1j−1 − p1j

p1j (t) = p1 exp[i(ωt − jk(lp + lb ))]. (25)

6.3. Experimental orders of magnitude of the wave oscillation parameters

You might also like